Troubleshooting on AWS using AI-powered tools

Your Amazon API Gateway endpoints start throwing intermittent errors. You do what most operations teams do: check Amazon CloudWatch, dig through logs, search the documentation. Hours pass. Still no root cause. You open a support case, and the AWS Support engineer resolves it quickly, then points you to a set of AI-powered tools that could have caught the problem before it escalated.

It's a familiar story. Most teams troubleshoot manually without realizing that AWS offers several AI-powered options, each approaching the problem from a different angle. Whatever the error and wherever you run into it, the right AI-powered tool can turn hours of troubleshooting into minutes.

This article covers four AI-powered troubleshooting tools. Each one works differently and fits a different moment in your workflow, from the AWS Management Console or terminal, troubleshooting a live incident, or working with AWS Support.

AI-powered troubleshooting options on AWS

Amazon Q Developer in the console

An AWS CloudFormation stack deployment fails, and the Events tab shows an error that you don’t recognize. Instead of opening documentation in a new tab and searching for the error code, you choose the "Diagnose with Amazon Q" button that appears next to the error. Amazon Q Developer performs multi-step reasoning across related services and logs, identifies the root cause, and provides step-by-step resolution instructions that are specific to your AWS account and resources.

Amazon Q Developer identifies that the AWS Identity and Access Management (IAM) role that's attached to the stack lacks s3:PutObject permissions on the target Amazon Simple Storage Service (Amazon S3) bucket. It explains why the deployment failed by tracing the relationship between the CloudFormation stack, the IAM role, and the Amazon S3 bucket policy. Then, it provides the exact policy statement that will resolve the error. Amazon Q Developer also links to the relevant IAM documentation. The entire diagnosis takes seconds and saves you hours compared with manually cross-referencing error codes with documentation.

Amazon Q Developer in the console offers two modalities. The first is an interactive chat panel where you can ask questions, get guidance, and troubleshoot interactively. The second is Diagnose with Amazon Q, an in-context button that appears next to errors across supported services. It performs multi-step reasoning with related resources and resolves errors. Both modalities cover a growing number of services and catch IAM permission errors across console pages.

Amazon Q Developer operates in read-only mode and uses the same permissions as your logged-in IAM identity. It requires the AmazonQDeveloperAccess managed policy or equivalent permissions. It adds no cost beyond your existing AWS access.

To get started, open the Amazon Q chat panel or look for the "Diagnose with Amazon Q" button when you encounter an error. For more information, see Asking Amazon Q to troubleshoot your resources.

Kiro CLI

A CI/CD pipeline fails with an error message that's buried deep in the build logs. Instead of manually scrolling through hundreds of lines, you open Kiro CLI in your terminal, enter the ModuleNotFoundError traceback, and ask it to find the root cause. Kiro CLI traces the cause to a missing dependency in your build configuration. It then shows you the exact line to update, explains why the conflict occurs, and generates a patch that you can apply with one command.

Kiro CLI is an AI-powered coding agent that runs entirely in the terminal. It analyzes errors, traces bugs across codebases, interprets logs, and generates fixes. It works with most programming languages, frameworks, or runtimes, making it useful for troubleshooting across your entire stack.

Kiro CLI supports custom agents for team-specific workflows. You can integrate Model Context Protocol (MCP) to connect to external tools or data sources and steering files to encode team coding standards. You can also use the ! prefix to run shell commands mid-conversation that send real output to the AI for analysis.

To install Kiro CLI and start a chat session, see the Kiro CLI documentation.

AWS DevOps Agent

You get a PagerDuty alert at 2 AM for elevated error rates on a payment service. AWS DevOps Agent picks it up automatically. It correlates CloudWatch metrics with a recent deployment in GitHub and identifies a code regression that introduced unhandled exceptions. Then it posts its findings to a Slack channel with a detailed mitigation plan. By the time the on-call engineer checks their phone, the root cause analysis is complete.

The posted finding includes a timeline that correlates the deployment at 1:47 AM with the error rate spike at 1:52 AM. It identifies the specific commit that introduced unhandled null references in the payment processing logic. It also links to the pull request and recommends rolling back to the previous version while the fix is prepared.

AWS DevOps Agent is an autonomous AI agent that automatically begins to investigate an issue the moment an alert comes in. It integrates with observability tools including CloudWatch, Dynatrace, Datadog, Splunk, New Relic, and Grafana. For CI/CD, it connects through GitHub, GitLab, and Azure DevOps and routes findings to collaboration tools such as Slack, ServiceNow, and PagerDuty. It also works across AWS, Azure, and on-premises environments through private MCP server connections.

AWS DevOps Agent goes beyond reactive incident response. It proactively analyzes patterns across historical incidents to generate prevention recommendations, targeting improvements in observability coverage, infrastructure optimization, deployment pipelines, and application resilience. Each investigation produces a detailed mitigation plan with step-by-step actions, not just a diagnosis.

Beyond investigations, AWS DevOps Agent serves as an always-available operational teammate. You can ask operational questions through natural language and get contextual answers that are aligned with your infrastructure. It also runs in headless mode, so you can interact with it directly from Kiro CLI.

To get started, set up your first Agent Space and try it on a recent incident. When needed, create an AWS Support case directly from an investigation, and provide Support engineers full context for a quick resolution. For more information, see Getting started with AWS DevOps Agent.

AI-enhanced troubleshooting in the Support Center

You’ve been using Amazon Q to troubleshoot a networking issue in the console. The suggestions helped narrow down the problem, but you need human expertise to resolve it. You open the AWS Support Center Console, describe the issue in natural language, and import your Amazon Q Developer conversation to preserve the troubleshooting context. The system recognizes the imported conversation about Amazon Virtual Private Cloud (Amazon VPC) connectivity issues. It asks targeted follow-up questions about security group rules and route table configurations. Instead of starting from scratch, the troubleshooting picks up exactly where Amazon Q Developer left off.

When you choose to escalate, the AI pre-populates a support case with the conversation history. The AWS Support engineer picks it up with full visibility into what you’ve already tried.

AI-enhanced troubleshooting replaces the traditional multi-field case creation form in the AWS Support Center Console. Instead of navigating dropdown menus and manually categorizing issues, you describe the problem in your own words. The system interprets the description, provides targeted troubleshooting guidance, and asks follow-up questions based on your specific AWS environment.

AI-enhanced troubleshooting is available with AWS Support plans. Basic plan users can access it for general questions and account and billing issues. Business Support+ and higher plans get full technical support. This feature is available only in the US East (N. Virginia), US East (Ohio), and Europe (Ireland) AWS Regions.

Conclusion

Operations teams that put these tools to work change how they handle errors. Amazon Q Developer becomes the first stop for everyday diagnosis in the console, and AWS DevOps Agent investigates production alerts on its own. Errors that used to take hours of triage get tracked down in minutes.

That speed comes from having the right tool at the right moment. Four entry points, four ways to work: the console, the terminal, a live production incident, or AWS Support. None of them rules out the others. You might open a problem in one tool and close it in another. Start with the one that fits where you already spend your day, and grow from there.

To learn more about how AWS Support plans and offerings can help you get the most out of your AWS environment, see AWS Support.

About the author

Henrique Santana is a Principal Cloud Support Engineer at AWS. He helps organizations modernize their technology stacks through container adoption and orchestration solutions. Henrique guides enterprises in overcoming containerization challenges and shares insights on cloud-native transformation journeys.